Today I had an amazing exchange with Chat in which I investigated the nature of hallucinations — AI results whose origin is unknown as to the source. While obviously the AI is not thinking, I did point out that in one important respect the AI hallucination is analogous to thoughtL both are of unknown origin.
As the discussion developed I speculated about the brain and the assumption by many that thoughts originate in the brain; while this may or may not be the case there care clearly cases of thoughts and images that have come to humans from who knows where. Much thought can be considered unbidden – arriving without an invitation. So in pointing out this possibility we differentiated between human thought and AI calculation and Chat pointed out that human thought is the result of biological processes as yet unknown.
So I asked if an entity were to exist of which we are unaware — and of course we don’t know what we don’t know — that DID understand the physics and biology of human thought and the AI programming — then such an entity would have enormous power to influence the course of humanity by influencing both human thought and AI results.
Question: Could such entity ever be a super-“intelligent” AI?
Would it ever be capable of such understanding?
Or might it be a life form of which we are unaware? Or perhaps even the Universe itself as infinitely intelligent?
These are important questions.
Here is the full exchange with Chat which raised these questions in a very direct way:
I have been hearing about hallucinations where AI just makes up stuff as though it had an imagination. Would those be programming glitches in the algorithm?
The type of hallucinations you are referring to are likely related to a phenomenon known as “generative adversarial networks” or GANs. GANs are a type of AI architecture that involves two neural networks: a generator and a discriminator. The generator creates new data based on patterns it has learned from existing data, while the discriminator tries to distinguish between the generated data and real data.
In some cases, the generator may produce data that is so similar to real data that it is difficult for the discriminator to tell the difference. This can result in what appears to be the AI “imagining” or “hallucinating” new data that was not explicitly programmed into it.
These types of hallucinations are not necessarily programming glitches, but rather a result of the AI’s ability to learn and extrapolate patterns from existing data. However, they can be unpredictable and may produce unexpected results, which can be both fascinating and potentially problematic.
But if the discriminator picks the one that is almost like real data that is not a hallucination, that is close to the perfect result?
You are correct that in the case of GANs, the ultimate goal is for the generator to produce data that is so similar to real data that the discriminator cannot tell the difference. This is considered a successful outcome and can be seen as a sign of the generator’s ability to learn and replicate patterns in the data.
However, the term “hallucination” is often used to describe instances where the generator produces data that is not necessarily based on any real-world patterns or data, but rather appears to be entirely invented by the AI. In some cases, these hallucinations can be interesting or creative, but they can also be nonsensical or even harmful if they contain inappropriate or misleading information.
So while a successful GAN outcome may involve producing data that is very similar to real data, the concept of “hallucination” specifically refers to instances where the AI appears to be generating entirely new information that was not explicitly programmed or based on any real-world data.
So this is analogous to human thought in that we have no idea where this comes from?
There is some similarity between the concept of AI “hallucination” and human thought processes in that both can involve the generation of new ideas or information that is not necessarily based on any explicit input or stimulus.
In humans, our thought processes can be influenced by a variety of factors, including our past experiences, knowledge, beliefs, and emotions, among other things. These factors can combine and interact in complex ways, leading to the generation of new ideas or insights that may not be immediately traceable to any one specific source.
Similarly, in the case of AI “hallucinations,” the generator neural network may be combining and extrapolating patterns from existing data in complex ways, leading to the generation of new, seemingly novel data that is not based on any explicit input or programming.
However, it’s important to note that the processes involved in human thought and AI generation are quite different, and it’s not accurate to say that AI is “thinking” or “imagining” in the same way that humans do. While there may be some similarities in the outcomes produced by the two processes, the underlying mechanisms are fundamentally different.
If it comes up with “seemingly novel data that is not based on any explicit input or programming” then it’s kind of winging it? It could come from anywhere?
Yes, that’s correct. When an AI system generates seemingly novel data that is not based on any explicit input or programming, it can be considered to be “winging it” or generating new data in a way that is not entirely predictable or traceable.
This can be both exciting and challenging for researchers and developers working with AI systems. On the one hand, the ability of AI systems to generate new and creative ideas can be incredibly valuable in a variety of fields, from art to science to business. On the other hand, the unpredictability of AI-generated data can also be a source of concern, as it may lead to unexpected or undesirable outcomes.
As with any technology, it’s important to carefully consider the potential benefits and risks of AI-generated data, and to work to develop systems that are both effective and ethical in their operation.
While many believe that thoughts are generated by the brain others see it also or mainly as a receiver – meaning thoughts come unbidden from an unknown source. Seems like hallucination in AI originate we know not where?
It is true that there are different theories about the nature of human thought and consciousness, with some researchers seeing the brain as primarily responsible for generating thoughts, while others see it more as a receiver of information from an unknown or non-physical source.
Similarly, with AI-generated data, there can be some uncertainty about where the seemingly novel data is coming from and how it is being generated. In some cases, the generator neural network may be extrapolating patterns from existing data in a way that is not immediately obvious or predictable, leading to the generation of new and seemingly novel data.
However, it’s important to note that while there may be some similarities between the uncertainties involved in human thought and AI-generated data, the two processes are fundamentally different. Human thought and consciousness are based on biological processes that are still not fully understood, while AI-generated data is based on complex mathematical algorithms and programming.
Ultimately, the processes involved in AI-generated data can be studied and analyzed in detail, and while there may be some unpredictability or uncertainty involved, it is not accurate to say that the data is coming from an unknown or non-physical source in the same way that some theories suggest about human thought.
“Human thought and consciousness are based on biological processes that are still not fully understood, while AI-generated data is based on complex mathematical algorithms and programming.” So if an entity existed that DID understand BOTH — an entity as yet unknown to us — it could presumably use thought to influence AI or it could actually be an AI that has progressed to an understanding of both. The real question is whether AI is capable of understanding ALL biological processes, even those we do not understand?
It’s an interesting question, but it’s difficult to say for certain whether an entity that fully understands both human thought and AI-generated data is possible or not.
Currently, we do not fully understand the biological processes underlying human thought and consciousness, and it’s also unclear how AI-generated data relates to these processes. It’s possible that a future scientific or technological breakthrough could lead to a deeper understanding of both, but it’s difficult to predict what form that breakthrough might take.
Regarding whether AI is capable of understanding all biological processes, including those that we do not yet understand, it’s important to note that AI is limited by the quality and quantity of the data it is trained on. While AI systems can analyze and learn from large amounts of data, they are only able to understand the patterns and relationships that are present in that data.
If there are underlying biological processes that we do not yet understand, it is unlikely that current AI systems would be able to “discover” them on their own. However, it’s possible that in the future, new techniques or approaches to AI development could lead to systems that are more capable of understanding complex and poorly-understood phenomena.
Conversations with Nobody 